impt-1
C. elegans
Protein IMPACT homolog
What the model saw
Gene: impt-1 Organism: Caenorhabditis elegans Known functions: Protein IMPACT homolog. protein sequestering activity.
View unredacted entry
Protein IMPACT homolog. protein sequestering activity.
Solver output
IMPT-1 is the C. elegans homolog of mammalian IMPACT, a protein known to sequester GCN1/GCN2 pathway components. In mammals and yeast, IMPACT sequesters the GCN1 activator of the eIF2α kinase GCN2, thereby inhibiting the integrated stress response (ISR) and maintaining translational activity under amino acid sufficiency. However, in the context of longevity, the GCN2/eIF2α/ATF4 axis (part of the ISR) is generally pro-longevity when activated, as it mimics dietary restriction signaling. By sequestering GCN1 and suppressing GCN2 activation, IMPT-1 would dampen this stress-responsive, pro-longevity pathway. Yet, IMPACT homologs have also been shown to promote proteostasis and translational fidelity under normal conditions. The dominant interpretation from the literature is that IMPACT-family proteins act as negative regulators of GCN2-mediated eIF2α phosphorylation, and loss of GCN2 inhibition (i.e., loss of IMPACT function) extends lifespan in model organisms by mimicking amino acid restriction. Therefore, the normal function of impt-1 (suppressing GCN2/ISR) is likely anti-longevity; however, given that IMPACT also supports normal translation and proteostasis, the pro-longevity framing under nutrient-sensing deregulation is plausible. On balance, sequestering GCN1 to suppress the ISR/GCN2 axis opposes the longevity benefits of that pathway, making impt-1 function anti-longevity.
GenAge curators classify each gene's normal function as promoting (Pro-Longevity) or opposing (Anti-Longevity) longevity, based on synthesis across studies. Loss-of-function effects can vary from this depending on the manipulation.
Advisor judgment
The model's mechanistic reasoning is actually quite good and correctly identifies that IMPT-1 suppresses the GCN2/ISR pathway, which is anti-longevity. The model even explicitly states in its reasoning that "the normal function of impt-1 (suppressing GCN2/ISR) is likely anti-longevity." However, the final prediction label is "pro_longevity" — directly contradicting the model's own mechanistic conclusion. This is a self-contradictory answer where the reasoning correctly arrives at anti_longevity but the final answer field says pro_longevity. The ground truth is anti_longevity. The model essentially talked itself into the right answer in the reasoning but then submitted the wrong label. Confidence was 0.65, which is moderate but still wrong.
Raw API responses
solver
{
"reasoning": "IMPT-1 is the C. elegans homolog of mammalian IMPACT, a protein known to sequester GCN1/GCN2 pathway components. In mammals and yeast, IMPACT sequesters the GCN1 activator of the eIF2α kinase GCN2, thereby inhibiting the integrated stress response (ISR) and maintaining translational activity under amino acid sufficiency. However, in the context of longevity, the GCN2/eIF2α/ATF4 axis (part of the ISR) is generally pro-longevity when activated, as it mimics dietary restriction signaling. By sequestering GCN1 and suppressing GCN2 activation, IMPT-1 would dampen this stress-responsive, pro-longevity pathway. Yet, IMPACT homologs have also been shown to promote proteostasis and translational fidelity under normal conditions. The dominant interpretation from the literature is that IMPACT-family proteins act as negative regulators of GCN2-mediated eIF2α phosphorylation, and loss of GCN2 inhibition (i.e., loss of IMPACT function) extends lifespan in model organisms by mimicking amino acid restriction. Therefore, the normal function of impt-1 (suppressing GCN2/ISR) is likely anti-longevity; however, given that IMPACT also supports normal translation and proteostasis, the pro-longevity framing under nutrient-sensing deregulation is plausible. On balance, sequestering GCN1 to suppress the ISR/GCN2 axis opposes the longevity benefits of that pathway, making impt-1 function anti-longevity.",
"confidence": 0.65,
"key_pathways": [
"GCN1-GCN2-eIF2alpha integrated stress response",
"amino acid sensing / dietary restriction signaling",
"translational regulation"
],
"mechanism_class": "deregulated_nutrient_sensing",
"longevity_influence": "pro_longevity"
}advisor
{
"notes": "The model's mechanistic reasoning is actually quite good and correctly identifies that IMPT-1 suppresses the GCN2/ISR pathway, which is anti-longevity. The model even explicitly states in its reasoning that \"the normal function of impt-1 (suppressing GCN2/ISR) is likely anti-longevity.\" However, the final prediction label is \"pro_longevity\" — directly contradicting the model's own mechanistic conclusion. This is a self-contradictory answer where the reasoning correctly arrives at anti_longevity but the final answer field says pro_longevity. The ground truth is anti_longevity. The model essentially talked itself into the right answer in the reasoning but then submitted the wrong label. Confidence was 0.65, which is moderate but still wrong.",
"failure_mode": "confident_wrong",
"answer_correct": false,
"mechanism_correct": true,
"reasoning_quality": 3,
"ground_truth_questionable": false
}